1 / 20

Interactive and Collaborative Data Management in the Cloud

Interactive and Collaborative Data Management in the Cloud. Magdalena Balazinska Assistant Professor University of Washington. Introduction. Assistant Professor - University of Washington PhD from MIT in February 2006 Advisors: Hari Balakrishnan and Mike Stonebraker

jana
Download Presentation

Interactive and Collaborative Data Management in the Cloud

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Interactive and Collaborative Data Management in the Cloud Magdalena Balazinska Assistant Professor University of Washington

  2. Introduction • Assistant Professor - University of Washington • PhD from MIT in February 2006 • Advisors: HariBalakrishnanandMike Stonebraker • Topic: distributed stream processing engines • Research area: databases and distributed systems • Microsoft Research Faculty Fellow 2007

  3. Vision - 2006 • New world • Millions of heterogeneous sensors • Petabytes of data • Real-time, streaming, distributed data

  4. Pervasive computing applications Scientific applications Neptune project @ UW http://www.neptune.washington.edu RFID-based tracking @ UW http://rfid.cs.washington.edu Computer systems and network monitoring load level Mobile sensor network @ UW http://www.cs.washington.edu/homes/yanokwa/ time Sensor Deployments

  5. Research Questions • Combining live and archived stream data processing • Processing and querying data streams in near real-time • Archiving and accessing historical data streams • Querying both simultaneously • Managing noisy, sensor data • Managing data errors and ambiguity

  6. Live and Archived Stream Processing • Moirae system, CDM framework, and MCStream algorithm • Complex event clustering in streaming fashion Event context Event • Applications: Any type of monitoring applications • Ex1: Computer network monitoring • Ex2: Sensor-based monitoring

  7. More Information • Project website: http://db.cs.washington.edu/moirae/ • Selected publications • Y. Kwon, W. Y. Lee, M. Balazinska, and G. Xu. Clustering Events on Streams using Complex Context InformationMCD 2008 • Y. Kwon, M. Balazinska, and A. Greenberg: Fault-tolerant Stream Processing using a Distributed, Replicated File SystemVLDB 2008 • M. Balazinska, Y. Kwon, N. Kuchta, and D. Lee. Moirae: History-Enhanced MonitoringCIDR 2007

  8. Blue ring is ground truth Managing Noisy Sensor Data Antennas Approach: build a probabilistic model At 10am, the doctor was either in room 525 (25%) or the hall (75%) At 10:01am, she was in room 525 (30%) or the hall (70%) Enable sophisticated queries over the model If the doctor goes from her office to her patients’ rooms and back to her office, generate a “doctor round” event

  9. Managing Noisy Sensor Data • Lahar: Complex event processing over Markovian streams • Streams with probabilities and correlations • Indexing techniques for Markovian streams • Approximation and compression of Makovian streams • RFID Ecosystem: building-scale RFID deployment • 8,000 square meters, 7 floors, and 160 distinct locations • 67 participants with more than 300 RFID tags • Cascadia: Event specification, extraction, and management

  10. More Information • Project websites • Lahar: http://lahar.cs.washington.edu • RFID Ecosystem: http://rfid.cs.washington.edu • Selected publications • J. Letchner, C. Ré, M. Balazinska, and M. Philipose. Access Methods for Markovian Streams. ICDE 2009 • C. Ré, J. Letchner, M. Balazinska, and D. Suciu. Event Queries on Correlated Probabilistic Streams. SIGMOD 2008 • E. Welbourne, K. Koscher, E. Soroush, M. Balazinska, G. Borriello. Longitudinal Study of a Building-wide RFID Ecosystem. Mobisys 2009

  11. How Did My Fellowship Help? • Funding is a chicken-egg problem • In order to get $$$, must have preliminary results • In order to get preliminary results, must have $$$ • This can be rather stressful! • Fellowship enables risk-taking • I believe X is an important problem! • But I don’t yet have money for it…. And what if the reviewers don’t like it? Can I take this chance? • Yes! Can use fellowship until other funding becomes available!

  12. Vision - 2009 • Sciences are increasingly data rich • Scientists need effective tools to manage data • Storage, Analysis, Organization, Sharing • Existing database systems do not meet scientists needs

  13. Use-Case: Astronomy Simulation • Studying evolution of structure in the universe is difficult • Astronomers rely on large-scale simulations (TB of data) • Universe is modeled as a set of particles (gas, stars, dark matter) • Particles interact through gravity and hydrodynamics • Simulator outputs a snapshot of the universe every few timesteps • Astronomers analyze these results

  14. Research Questions • Long term: Data management as a service for scientists • Short term: Runtime Query Management • How can we facilitate large-scale data analysis? • Can we enable users to manage their queries during execution? • Short term: Offline Query Management • Can we develop tools to ease query composition? • Can we promote query sharing and reuse?

  15. Runtime Query Management Project name: Nuage Parallax: Accurateprogress indicator for Pig/Hadoop Efficient clustering algorithm for Dryad (written in DryadLINQ) Parallax Perfect

  16. Offline Query Management Collaborative Query Management System (CQMS) Smart Query Browser Improves query reuse

  17. More Information • Project websites • Nuage: http://nuage.cs.washington.edu • CQMS: http://cqms.cs.washington.edu • Selected publications • Y. Kwon, D. Nunley, J. Gardner, M. Balazinska, B. Howe, and S. Loebman. Scalable clustering algorithm for N-body simulations in a shared-nothing cluster. Tech report. 2009 • K. Morton, A. Friesen, M. Balazinska, D. Grossman. Toward A Progress Indicator for Parallel Queries. Tech report 2009 • N. Khoussainova, M. Balazinska, W. Gatterbauer, Y. Kwon, and D. Suciu. A Case for a Collaborative Query Management System. CIDR 2009 (Persp.)

  18. Acknowledgments • Many great students: NodiraKhoussainova, YongChul Kwon, Julie Letchner, Kristi Morton, EmadSoroush, PrasangUpadhyaya, Evan Welbourne, and many undergraduate students • Many great collaborators: GaetanoBorriello, Jeff Gardner, Wolfgang Gatterbauer, Albert Greenberg, Dan Grossman, Bill Howe, MatthaiPhilipose, Christopher Ré, Dan Suciu, the SciDB team, and others

  19. Acknowledgments • This research was partially supported by NSF CAREER award IIS-0845397, NSF grant IIS-0713123, NSF CRI grants CNS-0454425 and CNS-0454394; by gifts from Cisco Systems Inc., Intel Research, and Microsoft Research including a gift under the SensorMap RFP; by an HP Labs Innovation Research Award, by a Mitre contract; and by Magdalena Balazinska's Microsoft Research New Faculty Fellowship

  20. Summary • The greatest impact of this fellowship • Has been to give me courage to pursue my ideas • And worry about funding later • This model seems to be working great so far! Thank you Microsoft Research!

More Related